Divergence 1
Running Head: Divergence
Divergence in Cultural Practices: Tastes as Signals of Identity
Jonah Berger, Chip Heath, and Ben Ho
Stanford University
Correspondence concerning this article should be addressed to Jonah Berger, Doctoral Student, Stanford University; Mail Code: 4840, 518 Memorial Way, Stanford, CA 94305. E-mail address: [email protected].
Divergence 2
Abstract
Divergence is pervasive in social life: people select different tastes to distinguish
themselves from others, and they abandon tastes when others adopt them. But while we know a
great deal about conformity, it predicts convergence, thus it has difficulty explaining why people
diverge. We propose an identity-signaling approach to divergence; people diverge to signal their
identity to others so that they can enjoy more fulfilling social interactions. Tastes gain signal
value through association with groups or types of individuals, but become diluted when members
of more than one type hold them. Consequently, different types of people will diverge in the
tastes they select, and they will abandon tastes that are adopted by members of other social types.
(117 words)
Divergence 3
The young always have the same problem--how to rebel and conform at the same time. They solve this problem by defying their parents and copying one another. - Quentin Crisp
Kids want to show they are different from their parents and jocks want to separate
themselves from geeks. Shanghai residents avoid purchasing Volkswagen Santanas because they
“are a favorite first car among the nouveaux riches outside the big cities” (Wonacott, 2004, p.
B1) and Manhattanites wore mesh trucker hats until the bridge-and-tunnel crowd adopted them
(Barker, 2004). Intellectuals want to show they are more thoughtful than the masses and the
original members of any cultural scene (i.e. music, style, philosophical schools, etc.) want to
differentiate themselves from the poseurs that come along later. The social process that
underlies all these examples is one of divergence. People select tastes (e.g., possessions,
attitudes, behaviors, etc.) that distinguish them from other types of people, and they abandon
tastes when the wrong types of people adopt them. Divergence is pervasive in social life.
These facts are so obvious they hardly need stating save one important observation: our
theories in social science are generally unable to account for them. One of the most widely
discussed principles in social psychology is conformity (Allen, 1965; Allport, 1924; Asch, 1951;
Deutsch & Gerard, 1955; Festinger, 1950; Graham, 1962; Milgram, Bickman, & Berkowitz,
1969; Moore, 1921; Sherif, 1936; Stafford, 1966; Venkatesan, 1966; Willis, 1965); indeed
conformity has been raised to the status of a social law (see Social Impact Theory, Latane, 1981).
Similar tendencies have been recognized under different names by researchers in sociology (e.g.,
mimetic isomorphism, DiMaggio & Powell, 1983; Meyer & Rowan, 1977; Scott, 1995) and
economics (bandwagon effects, Liebenstein, 1950; Granovetter & Soong, 1986, herding behavior
and information cascades, Banerjee, 1992; Bikhchandani, Hirshleifer, & Welch, 1992).
Divergence 4
While conformity processes are obviously important, they cannot account for a world in
which people select tastes that distinguish themselves from others and abandon tastes that are
adopted by others. Indeed, models of the diffusion of innovations and cultural practices, which
are based implicitly on conformity dynamics (Bass, 1969; Rogers, 1983), are well equipped to
explain increases in adoption, but as Macy and Strang (2001) note, they are “poorly equipped to
account for almost anything else” (p. 148). We know that people diverge but why?
The uniqueness literature in psychology (Fromkin, 1970; Snyder & Fromkin, 1980) has
made the important observation that people experience drives to differentiate as well as to
belong, but this literature is much smaller than the literature on conformity, and it leaves some
important questions unanswered. It argues that individuals feel badly in situations where they
are overly similar to others, and they repair the negative feelings by emphasizing aspects that
make them different from others. By focusing on the feelings and reactions of individuals,
however, the uniqueness literature is not equipped to explain the social character of divergence--
the uniqueness literature would allow individuals to diverge in unique, perhaps bizarre,
directions, but we suggest that most divergence is driven by a process that unfolds at the social
level.
We propose an identity signaling approach to divergence; people diverge to signal their
identity to others so that they can enjoy more fulfilling social interactions. Tastes and practices -
- wearing hiking boots or Manolo Blahniks, dropping an apposite reference to Kurt Lewin or
Wittengenstein, or mentioning an obscure foreign film, independent band, or poet-- signal
people’s internal and complex characteristics to others. When different types of people diverge
in the tastes they select, individuals will find it easier to identify a given other’s type, making it
easier to find people who are similar (to ensure future interactions with them), and to identify
Divergence 5
people who are not similar (making those interactions more beneficial as well). Wearing hiking
boots allows hikers to congregate, but it also allows non-hikers to know whom to consult on the
occasion they decide to go on a hike. Signals, by their nature, must be social--people can only
send signals that are understood by others. But because they are social, signals can sometimes be
appropriated improperly by onlookers who really aren’t similar. The unhip might want to be
considered hip, the middle class to be treated as wealthy, and the bridge-and-tunnel crowd to be
confused for Manhattanites. When these poseurs adopt another group’s taste, that taste loses its
ability to signal identity effectively. This may lead to another type of divergence: the original
taste holders may abandon the taste in favor of a new, undiluted signal of identity.
Before presenting our theory, we first review previous research that relates to divergence,
and then present three key questions that are unresolved by previous literature: (1) Why is
divergence not idiosyncratic? Collections of people often diverge in the same direction at the
same time, an observation that is difficult to explain within the traditional uniqueness literature.
(2) Where do people diverge? People care about divergence much more in some domains than
others, about sharing the same music tastes than the same bike lights. We should understand
why domains vary in their use as signals of identity. (3) Who drives divergence? People often
abandon tastes when other types of people adopt their practices. Abandonment is a particularly
costly form of divergence because it requires people to give up a taste they once favored, and we
should understand why it occurs. The body of the paper addresses each of these questions in
turn, providing a simple mathematical model that clarifies our predictions, and then reviewing
evidence consistent with the model from previous work in psychology, sociology, economics,
and consumer behavior as well as our own research.
Divergence 6
Previous Literature
Divergence driven by a desire for uniqueness
Although we know much more about conformity than divergence, a small number of
researchers have also considered principles that may lead people to diverge. However, as we
discuss below, their theories still leave important questions unresolved.
Social scientists have argued that individuals want to be different. Early on, economic
theorizing discussed the drive for difference, suggesting people choose things that set them apart
from the masses (Veblen, 1899 [1912]). Liebenstein (1950) argues people “search for
exclusiveness…through the purchase of distinctive clothing, foods, automobiles, houses, or
anything else that individuals may believe will in some way set them off from the mass of
mankind” (p.184), and discusses “snob effects” or cases where individual demand is decreased
because others are consuming (also see Liebenstein, 1976; Granovetter & Soong, 1986).
Psychologists have also made a concerted effort to study the drive for uniqueness
(Fromkin, 1970; Snyder & Fromkin, 1980). They suggest people are “motivated to seek some
sense of difference relative to others” and “when a person feels a very high degree of similarity
relative to another, a negative emotional reaction should occur” (Snyder & Fromkin, 1984, p. 34)
driving the person to take steps to decrease perceived similarity. When people feel too similar to
others, they focus on differences between themselves and others who seem similar (Ganster,
McCuddy, & Fromkin, 1977), misremember actual levels of similarity (Byrne and Griffitt, 1969;
Smith, 1975; Snyder & Batson, 1974), generate more unique uses for ordinary objects (Fromkin,
1968), and place higher value on scarce future options (Fromkin, 1970). Literature on the pursuit
of distinction, then, would predict individuals might diverge in their tastes to maintain a unique
sense of self.
Divergence 7
The uniqueness literature obviously has to grapple with explaining why a large number of
studies have found conformity. Its solution has been to suggest that, while extreme similarity is
a negative state, moderate similarity produces a positive state and people tend to conform to
others when they feel only moderately similar. It suggests, however, that if conformity
researchers had confronted subjects with a more extreme degree of similarity, pressures to be
unique would have intensified, and the studies would have found divergence.
Though another stream of literature--optimal distinctiveness (Brewer, 1991, 1993;
Brewer & Pickett, 1999, 2002; Brewer & Roccas, 2001)--does not study divergence directly, it
suggests people can resolve countervailing drives for assimilation and differentiation by
emphasizing different social identities. Here, “social identities are selected and activated to the
extent that these identities help achieve a balance between needs for inclusion and for
differentiation” (Brewer & Pickett, 1999, p. 73). When people feel overly similar, their renewed
need for individuation drives them to emphasize their distinctive group memberships (e.g., band
member rather than Plainsville High student); when people feel excessively different, they
emphasize their broad, generic social category memberships (e.g., Plainsville High student rather
than chess club member). People can also satisfy their differentiation motive through
differentiating larger groups into subgroups with which to identify, or “making intergroup
comparisons (i.e. focusing on distinctions between an ingroup and contrasting outgroups),”
(Brewer & Pickett, 1999, p. 74). Membership in groups allows people to feel similar and
different at the same time: similar because they are part of a group and different because the
group is separate from the masses.
Optimal distinctiveness is extremely useful in pointing out a way people can resolve
competing drives for assimilation and distinction, but the main mechanism that has been studied
Divergence 8
by this research is internal. People activate or mentally emphasize different group memberships
depending on whether they feel overly similar or overly distinctive. The uniqueness literature
stresses similar internal tactics by saying people can achieve uniqueness by mentally
emphasizing ways in which they are different from others (e.g., Ganster, McCuddy, & Fromkin,
1977) or distorting similarity information (e.g., Byrne and Griffitt, 1969; Smith, 1975; Snyder &
Batson, 1974). Internal processes, however, seem too flexible to account for the costly behaviors
we observe when people diverge behaviorally from others in the social environment. Why get a
body piercing when you can mentally emphasize your membership in club culture? Why blow
big bucks on a Prada handbag when you can mentally tally the value of your stock portfolio?
Another limitation of previous research on divergence is that it has focused on states that
seem highly transient. The uniqueness literature argues divergence will occur when “a person
feels a very high degree of similarity relative to another,” (Snyder & Fromkin, p. 34). Optimal
distinctiveness suggests that individuals will choose to emphasize more divergent identities when
they feel depersonalized. Transient states cannot explain why tastes die or only come back in
style after long periods. Sure, people might abandon tastes if they are in a state in which they
feel overly similar or depersonalized, but why would these individuals not re-adopt those tastes
once the transient state passed? A leisure suit fan might stop wearing a leisure suit after visiting
a club filled with leisure-suited men, but put it on again after a particularly isolated day at work.
An English professor might stop talking about Derrida at a party where everyone else is doing
so, but take it up again at the next one. We commonly think of fads as transient, but their
transience takes place on time scales -- weeks, months, or years-- that are mismatched to the
momentary processes that have been studied in the lab.
Divergence 9
Questions Raised by Prior Approaches
Prior approaches provide some important insight into divergence, but they leave a few
important questions unresolved.
Question 1: Why isn’t divergence idiosyncratic?
Perhaps the most important limitation of previous theories is that they have primarily
focused on individual level analysis. The uniqueness literature argues divergence will occur
when “a person feels a very high degree of similarity relative to another,” (Snyder & Fromkin,
1980, p. 34). Optimal distinctiveness also focuses on the individual, suggesting “individuals
will…define themselves in terms of social identities that are optimally distinctive,” (Brewer &
Pickett, 2002, p. 258). Because these theories have focused on individuals, they have difficulty
explaining why Group A diverges from Group B, or why everyone in a group diverges in a
similar direction, often at the same time.
We note here that there is another, quite different, class of theories that would predict
idiosyncratic divergence. Many of the taste domains we will discuss in this paper may be
affected by internal drives for novelty and variety (Acker & McReynolds, 1967; Cattell, 1975;
Farley & Farley, 1967; McAlister, 1982; Kahn, 1995). Novelty or variety-seeking would suggest
people diverge when they get bored with tastes, but this individual focus means they also have
difficulty explaining non-idiosyncratic divergence.
Question 2: Where do people diverge?
The notion of moderate similarity also doesn’t specify where people should diverge.
Momentary thought suggests there are certain domains where people are much less willing to
accept similarity. People get more upset when they wear the same dress to a party than when
they bring the same toothbrush to a camping trip. Why?
Divergence 10
The uniqueness literature has little to say about how different choice domains induce
more or less divergence. Most of this literature has focused on personality and individual
differences in the need for uniqueness (Snyder & Fromkin, 1977; Tepper, 1994; Lyne & Harris,
1997; Tian, Bearden, & Hunter, 2001). Consequently, we know that certain individuals may
prefer unique products, but very little about how the situation itself, or the domain in which the
choice is made, influences the degree of divergence preferred. By focusing on individuals rather
than situations, these approaches ignore the fact that across individuals, people tend to diverge
more in some domains rather than others (e.g., music choices or hairstyles rather than computers
or stereos). Understanding why people tend to diverge in certain domains provides deeper
insight into what drives divergence in the first place.
Question 3: Who drives divergence?
The uniqueness literature argues people should be most likely to diverge when they feel
extremely similar. This leads to some curious predictions. Because an individual is likely to be
highly similar to their friends, the prediction is that punks should be more likely to diverge from
other punks than from mainstream culture. On a macro level, people in LA shouldn’t worry
much that their cultural practices are being adopted by people in Des Moines (who differ on
many dimensions and thus are only moderately similar); rather, they should worry most about
other LA residents. We suggest these predictions are false and that understanding divergence
will require us to think in a more complex way about the social process by which one type of
people tries to distinguish itself from other types.
Summary: Three Questions
Taken together, previous work has provided some insight into why people may diverge,
but cannot explain many cases of divergence and leaves some important questions unanswered.
Divergence 11
Previous literature on uniqueness and distinctiveness has added a great deal to a literature which
has predominantly focused on processes of conformity and convergence. But we think there is
additional insight to be gained by emphasizing the social process by which identity is
communicated.
Tastes as Signals of Identity
Our consumption habits are now a form of social currency. The man who uses Callaway golf clubs, drives a Jaguar, and wears Ralph Lauren apparel…is a man separate and apart from the man who uses a Penn fishing reel, drives a Dodge Durango, and wears Levi’s (Vincent, 2002, p. 11)
To address the three questions raised in the previous section we propose an identity
signaling approach to taste divergence. Though signaling identity through “one’s attitudes and
preferences is presumably a ubiquitous social communication process, only limited research in
psychology has investigated its characteristics.” (Shavitt & Nelson, 1999, p. 38).
Signals are compact indicators of information that is hard to observe or hard to
summarize. When first meeting a group of people, for instance, we may want to know “which of
these people is the type of person I would enjoy talking to?” We often use signals to determine
which people we want to interact with further. Such signals may include the person’s tastes, e.g.,
the clothes they wear, the attitudes they profess, or the phrases they use. A person’s tastes
provide insight into his/her identity or social type: in China, a Volkswagen Santana driver is
likely a new money suburbanite. Signals can be extremely easy to detect, like whether someone
has a pierced nose, or more difficult, like whether they correctly use phrases like hegemonic
discourse, economic value added, or self-regulatory process.
Our identity-signaling approach suggests people diverge to signal their identity to others
so that they can enjoy more fulfilling social interactions. Tastes gain signaling value through
their association with groups or types of individuals, but these signals are only valuable to the
Divergence 12
degree that they actually provide information. If different types of people send an identical
signal, it will have no value in distinguishing between types. If Manhattan residents and
weekend wanna-bees both wear mesh trucker hats, then an observer won’t be able to determine
whether a hat wearer has information about the newest underground band or hip night spot.
Divergence makes it easier for people to identify others who are similar (to ensure future
interactions with them), and different (to interact appropriately with them as well--e.g., avoiding
disliked types or deferring to expert types). We suggest different types will diverge in the tastes
they select and they will abandon tastes when their meaning becomes diluted.
Our model assumes that both conformity and uniqueness processes are important, but the
focus of the model is collective rather than purely individual; we focus on how meaning is
constructed and destroyed as collections of people interact at a social level. Individuals converge
with similar others to imbue signals with meaning, and they diverge when their signals are
adopted by the wrong types of people. Signals are constructed at a collective, social level, and
are diluted through a collective, social process.
Introducing a Model of Identity Signaling
We propose a simple model that can account for much of the observed divergence. The
purpose of the model is to rigorously articulate our assumptions and lay bare how our proposed
mechanism works. Though the mathematics describing the model have been fully derived, we
limit the mathematical details to the appendix. The major concepts of the model can be readily
understood without them.
To illustrate the ideas of the model, we will construct it in three stages. In Model 1, we
show how different types of people can use signals to improve their social interactions. Model 2
considers in which domains people will choose to signal. Finally, Model 3 adds a poseur who
Divergence 13
would like to be treated as another type and poaches the signals of that type; we show how this
kind of poaching behavior may dilute the value of signals and lead individuals to abandon tastes
and behaviors they previously held.
Why Isn’t Divergence Idiosyncratic? Signaling Identity
Previous literature notes that tastes can have symbolic value (Abelson & Prentice, 1986;
Herek, 1986; Jetten, Branscombe, Schmitt, & Spears, 2001; Levy, 1959; Park, Jaworski, &
MacInnis, 1986; Prentice, 1987; Shavitt, 1990; Solomon, 1983; Zaltman & Wallendorf, 1979),
but has given less attention to how tastes actually gain this symbolic meaning (see Davis, 1992).
Research suggests people express their identity through their attitudes and possessions (see
Sirgy, 1982 for an early review; Abelson & Prentice, 1986; Ball & Tasaki, 1992; Belk, 1988;
Escalas & Bettman, 2003; Holman, 1981; Kleine, Kleine, & Kerman, 1993; Levy, 1959;
Rubinstein, 1995; Wicklund & Gollwitzer, 1982) and other work suggests that tastes act as part
of the social communication system, communicating aspects of the self to others (e.g., Davis,
1992; Douglas & Isherwood, 1978; Holman, 1981; Holt, 1998; McCracken, 1986; Richins,
1994a, 1994b; Solomon, 1983; Schlenker, 1980). But for individuals to be able to communicate
or signal identity through tastes, others must be able to decipher or understand those signals and
previous literature has not systematically described how signals acquire or change their meaning.
Meaning Construction: Signals are socially constructed
Signals are valuable to the degree that they provide information, but individuals cannot
construct signals on their own.
Coordination Hypothesis: tastes gain social meaning as signals when they are expressed by similar types of individuals
We use the term “types” rather than groups because tastes frequently carry meaning even when
they span traditional demographic groups. People who wear their baseball caps backwards or
Divergence 14
who talk about Foucault may come from different races, ages, and cultures. But, as long as they
are similar in some way (e.g., they all listen to the same music or have the same disdain for
reifying power relationships), the taste has meaning. Thus type refers to some degree of
similarity between individuals expressing the taste.
Signals cannot be constructed solely by individuals because idiosyncratic displays could
not be understood by others. Consider the inference problem faced by others when an individual
sets out to convey a distinctive identity by choosing a unique combination of markers (e.g.,
wearing a pith helmet and a football jersey): no one will know what that combination stands for
and it will have no expressive value. Thus, a person’s decision to adopt or abandon a taste is not
just a personal choice made to create a self-image. Tastes signal identity to the broader social
world.
More effective signals can be constructed when individuals join with similar others,
conjointly displaying the same signal. If a group of lacrosse fans or mushroom hunters all adopt
the pith helmet and football jersey combination, people may start to recognize it as a signal of
identity as opposed to simply being bizarre.
Model 1: Signaling Identity
The model we develop uses the standard logic of game theory. Although we will be
using the language of choice, we are not arguing that types have agency; instead we are
assuming that certain “picks” are more likely to be stable in an evolutionary process of variation
and selection. That is why game theory has been useful in biology where selection of strategies
by animals are not conscious.
Imagine a community where actors differ in their preferences, i.e. their interests and
tastes. For the purposes of simplicity, let there be two types of actors (Type A and Type B)
Divergence 15
where all actors of the same type have identical preferences. Actors play a game with two
stages. In the first stage, each actor chooses a signal. Every actor can signal by investing in
some activity within a fixed domain. In the clothing domain, for instance, an actor can select the
color or style of his jacket. In the domain of musical taste, an actor could develop knowledge of
independent bands or opera.
In the second stage of the model, the actor randomly meets another actor that she has
never met before. She does not observe the other actor’s type directly, but observes the taste the
other has signaled. Signals are such that they can observed relatively quickly (e.g., during a brief
conversation). The receiver observes the sender’s signal and then decides how she wants to treat
the sender in their interaction. Suppose A is an outcast and B is an elite. If the outcast type is
happy to interact with both As and Bs, but the elite only wants to interact with other Bs, then A’s
would pursue interaction with people who signaled membership in either type while Bs would
only pursue interaction with people whose signals indicated they were Type Bs.
As a base case, let us assume that individuals want to be treated as they see themselves.
People prefer to be treated as members of their type no matter whether they are interacting with
members of their own type or members of another type.
Understanding the Model
In this simple model, there is no conflict of interest; everyone wants to treat senders
exactly how the senders wish to be treated. This model predicts that the two types will
coordinate on different signals so that members of each type can uniquely signal their identity.
Type A actors should coordinate on one signal, i, and Type B actors should coordinate on some
other signal j. These signals are social in two ways: (1) they are coordinated among members of
the same type who choose the same signal, but also (2) they are coordinated across different
Divergence 16
types who must choose different signals. This coordination improves interactions both with
people of the same type and with people of other types.
Experimental results support our proposition that tastes gain meaning through expression
by similar types of individuals, not by idiosyncratic individuals acting on their own (Berger &
Heath, 2005b). Stanford undergraduates were given information about one or four users of a
novel brand and were asked to rate the brand on a number of personality dimensions, e.g.,
whether it was rugged, calm, reliable, exciting, etc. (see Aaker, 1997). Half were given
information about one brand user (a 25 year-old recent graduate who likes surfing, etc.) and the
other half was also given information about three additional users (all around the same age, who
all liked some extreme sport). We have argued that signals will be clearer when they are
expressed in a coordinated way by multiple individuals. Indeed, although participants in the
similar type condition came up with remarkably similar perceptions of the brand (α = .94), there
was significantly lower congruence about the brand when it was used by only one individual (α =
.64, t(20) = 3.45, p < .001, Duhachek & Iacobucci, 2004).
Seeing expression as socially coordinated differs heavily from the traditional perspective
on self-expression which has emphasized the self who is doing the expressing. Traditionally the
literature has argued that individuals use possessions, attitudes, and beliefs to pursue unique self-
image (Fromkin, 1978; Gross & Osterman, 1971; Liebenstein, 1950), “maintain self-concept”
(e.g., Ball & Tasaki, 1992, p. 155; Sirgy, 1982) and “express self-identities” (e.g., Prentice, 1987,
p. 993) and personal values (e.g., Herek, 1986; Katz, 1960). Escalas and Bettman (2003) argue
individuals “engage in consumption behavior to construct their self-concept and create their
personal identity” (p. 339, see also Miller, McIntyre, & Mantrala, 1993). Similarly, the
functional approach to attitudes (Herek, 1986; Katz, 1960; Smith, 1947; Smith, Bruner, & White,
Divergence 17
1956) focuses on psychological benefits individuals derive from expressing “attitudes
appropriate to his personal values and to his concept of himself,” (Katz, 1960, p. 170).
While some researchers do acknowledge the social and symbolic function of attitudes
(e.g., Abelson & Prentice, 1986; Shavitt, 1990), this “social identity function of attitudes…has
primarily been investigated from the perspective of the attitude holder,” (Shavitt & Nelson, 1999,
p.38). More generally, Brewer (1991) notes that “ in recent years, social psychologists have
become increasingly ‘self’-centered,” (p. 475) and that “until recently, social psychological
theories of the self focused on the individuated self-concept, a person’s sense of unique identity
differentiated from others,” (Brewer & Pickett, 1999, p. 71). In this literature, individuals
maintain self-esteem by expressing the individual identity they themselves wish to create.
By focusing on individuals, however, the literature has neglected the fact that individuals
can’t express meanings that are not socially understood. Adolescent boys can purchase muscle
cars to bolster their masculine self-concept, but they cannot define the meaning of that symbol.
In fact, that symbol may lose meaning over time when muscle cars are purchased by kids who
can barely shave. Individuals also cannot control shifts in the meaning of the signals they send.
People can buy SUVs because they want to be seen as outdoorsy, but if society comes to see
SUVs as silly gas-guzzlers, the wanna-be outdoorsmen are out of luck. Finally, by concentrating
on how individuals construct identities, the previous literature has neglected the broader question
of how tastes gain meaning in the first place. Previous research has not entirely ignored the fact
that tastes have symbolic meaning, it has just not systematically discussed how meaning is
acquired.
We emphasize the social nature of identities and attempt to understand how tastes acquire
their social meaning as signals of social identity. Our approach finds theoretical support from a
Divergence 18
diverse set of literatures. Work on social identity (Tajfel & Turner, 1986) also takes identity as
derived from outside the individual, but this literature focuses primarily on how individuals use
social identities to manage self-esteem and how individuals react when they adopt a particular
identity (e.g., by venerating the ingroup and favoring it in resource allocations). The theoretical
approaches of symbolic interactionism (Blumer, 1969; Mead, 1934; Solomon, 1983) and
poststructuralism (e.g., Holt, 1997) also note the social nature of the self and recognize symbols
acquire shared meaning through socialization, but they do not make clear how this process
occurs.
The notion that cultural tastes demarcate social groups is familiar to sociologists
(Bourdieu, 1979; DiMaggio, 1987; Douglas & Isherwood, 1978; Simmel, 1904; Goffman, 1951),
but they often take these markers as given. As Douglas and Isherwood (1978) note, though
tastes themselves “are neutral” (p. 12) and we argue that it is only through expression by
multiple individuals of the same type that tastes are imbued with meaning.
Summary: Why Isn’t Divergence Idiosyncratic?
To summarize, our model suggests that tastes facilitate interaction and acquire shared
meaning through association with individuals of a certain type.
Seeing tastes as signals of (type) identity immediately addresses the first question
regarding moderate similarity, namely why divergence typically isn’t idiosyncratic.
Idiosyncratic divergence is difficult to understand. When people diverge with similar others,
however, they communicate an understandable social identity. We now move to examine how
our signaling approach can resolve another question neglected by previous approaches, namely
where does divergence occur?
Divergence 19
Where Do People Diverge? The Question of Domains
Simple observation shows that there is more divergence in the clothes people wear and
the cars they drive than in the notebooks they write in or bike lights they use. Similarly, people
care a lot more when their music tastes are co-opted by the mainstream than when everyone else
uses a similar type of umbrella. What predicts the domains where people are more likely to
diverge?
The Signaling Approach to Taste Domains
Little research has examined why divergence might be greater in certain domains but
previous research has provided some suggestions. Divergence may increase in publicly visible
domains. Ratner and Kahn (2002) suggest self-presentational concerns lead people to choose
more variety in public settings than private ones (see also Bearden & Etzel, 1982).
Divergence may also increase in domains that are personally important. Campbell (1986)
found that people were more likely to make judgments that protected their self-concept (e.g.,
underestimating the number of other who share their abilities) in personally important areas.
Similarly, Kernis (1984) found that the need for uniqueness influenced false-consensus bias on
various trait estimates, but only for people who listed a particular trait as highly important to
their self-schemas. This work suggests people want to be more unique in areas they find
personally important; a coin collector should care a lot about having unique coins.
These approaches, however, cannot explain many cases of divergence. Public domains
may increase the drive for divergence, but people often seek divergence on practices (e.g., music
tastes or home furnishings) that are not particularly public. Similarly, personal importance
cannot explain why certain domains (e.g., music tastes or clothing) seem to elicit more
divergence across people.
Divergence 20
At an idiosyncratic personal level, if people want to protect their self-concepts, then
people who are self-schematic about a particular dimension should try harder to diverge from
others on practices relevant to that dimension. A coin-collector and bottle-cap collector both
probably seek to be unique in their own particular hobby domains. But at a broader social level,
individuals cannot idiosyncratically select domains to express their identity if they want others to
receive their signal. Signaling domains must be coordinated socially, and domains people use to
express identity should overlap strongly with the domains that others use to infer identity.
Domain Overlap Hypothesis: Because of the coordination required for signals to be understood, people should choose to self-express in domains others use to infer identity.
Berger & Heath (2005a) found experimental support for this hypothesis in a broad Internet
sample from across the United States (mean age = 38, 82% Caucasian). Separate sets of raters
were given a variety of taste domains (e.g., hairstyle, toothpaste, jacket, etc.) and rated each on
either self-expression (how much choice in that domain contributes to self expression) or identity
inferences (how much choice in that domain is used to make inferences about others). Raters
exhibited high agreement about which domains were used in self-expression (α = .93) and about
which domains they used to make inferences about others (α = 91). Supporting the Overlap
hypothesis self-expression and inference making were highly correlated (r = .97). In general,
respondents suggested they were more likely to infer identity in domains like hairstyle or musical
tastes than backpack or stereo brand. These results support the proposition that self-expression is
not an idiosyncratic activity; there is coordination on what domains people use to infer things
about others and people express themselves in these identity-related domains.
We suggest divergence is driven by the need to maintain clear signals of identity, thus
divergence should be greater in identity-related domains.
Divergence 21
Domain Divergence Hypothesis: Divergence will be greater in identity-related domains.
Applying this hypothesis to the divergent findings of convergence and divergence proves
immediately useful. While many conformity studies used choices likely to have a right answer
and hence choice was not particularly self-expressive (e.g., length of a line, Asch, 1951,
movement of a point of light, Sherif, 1936, or weight of items, Allport, 1924), the few studies
where divergence was found concern categories such as attitudes (Weir, 1971) that relate more to
identity.
To test the domain divergence hypothesis, Berger and Heath (2005a) asked a broad
Internet sample from across the United States to choose options in various preference domains
(e.g., dish soap, stereos, hairstyles, favorite CD). In each domain, they were told 65% of people
preferred Option A, 25% preferred Option B, and 10% preferred Option C, and they indicated
which option they would choose. Supporting the Domain Divergence hypothesis, people
preferred more divergence in identity related domains (Figure 1); respondents were more likely
to select the option preferred by 10% of others in domains like hairstyle and favorite music artist
than dish soap or toothpaste. People also avoided tastes held by too many others in identity-
related domains; while only 24% of people chose Option A in domains used to infer identity
(median split), 46% did so in less identity-related domains.
These results suggest that divergence is more likely in domains used to infer identity.
But this raises another important question. Why are some domains used to infer identity? Why
are hairstyles or music (as opposed to back packs or bike lights) better signals of identity?
Previous research suggests visibility or publicness may influence the degree to which
tastes are used to make inferences about the individual (Belk, 1991; Ratner & Kahn, 2002;
Shavitt & Nelson, 1999). But the public/private distinction does not capture enough of the
Divergence 22
variance (see Bearden & Etzel, 1982). Indeed, in our own 65%/25%/10% study, domain
publicness was somewhat predictive of divergence, but only marginally so. A moment’s
introspection will quickly suggest that domain visibility cannot be the only factor that influences
identity inference making. A person’s clothing is about as observable as their notebook, yet only
one of those domains is used to infer identity. And we often infer identity from domains that
aren’t particularly observable (e.g., music tastes aren’t easy to observe in most situations). Thus,
while visibility is likely to be important, it is not the whole story.
Functionality
We suggest that some domains operate more effectively as signals of identity because
they are easier to interpret. Most cultural practices have some functional value as well as some
self-expressive or symbolic value (Abelson & Prentice, 1986; Levy, 1959; Zaltman and
Wallendorf, 1979). We argue that practices with a salient functional component will be harder to
interpret as identity signals. Did someone buy the new power-sander because its design appealed
to their aesthetic sense or because they had something to sand? When options are functionally
equivalent (e.g., a black jacket vs. a brown jacket), it is less likely that an individual’s choice will
be attributed to function.
Based on standard theories of psychological discounting (Kelly, 1973; Morris & Larrick,
1995), we predict that people will have greater difficulty identifying a practice as signaling
identity when it has a higher functional component. Thus, we predict that certain taste domains
may be particularly suited for signaling identity because others cannot easily attribute choice
within the domain to functional reasons.
Afunctionality Hypothesis: Domains will operate more effectively as signals of identity when choice in them is less attributable to function.
Divergence 23
Bike lights are less likely to signal identity than hairdos because the choice of a bike light is
based more on function. Similarly, in cultural beliefs and practices where people think there is a
right answer, functionality is high, so those beliefs are less likely to be used to infer identity.
Admiring Newton as a scientist will be less identity revealing than admiring Rousseau as a
philosopher. Worshiping a particular god will be less identity revealing in a monotheistic culture
(where there seems to be a single right answer) than in a polytheistic culture.
Model 2: Introduction of Costs and Influence of Domain on Signal Interpretation
In the previous model we did not differentiate among domains. We now expand the
model to allow different domains to provide different benefits. This allows us to ask the
following question: If we allowed types to choose which domain to signal in—hairstyle, music
taste, clothing—which domain would they pick?
Formally, assume that an activity, i, provides benefits, bi and costs, ci,x, to the actor, x. In
the domain of fashion, the activity of wearing a biker jacket provides the benefit of warmth and
protection but imposes the cost of the purchase of the jacket. In the domain of musical tastes,
specialized knowledge about hip-hop imposes the cost of time spent learning who was featured
on a given artists last album. If the domain is a certain kind of lingo, the cost is time spent
listening to others speak the lingo. For simplicity, assume each actor can only invest in one
signal each period.
Understanding the Model
The main insight provided by Model 2 is that signaling is more likely to occur in
afunctional domains. In functional domains, most people will select the option that provides the
most benefit and consequently choice is not a good signal of identity. Choices on intellective
tasks or choices of stereos or notebooks will probably not provide good signals. Some practices
Divergence 24
can also be broken down into aspects that might signal identity and those that don’t. On a cold
day, the act of wearing a coat doesn’t provide a signal but the coat’s color does provide a signal
because the benefits of various colors are similar. Thus, we expect signaling is more likely in
afunctional domains.
Consistent with the afunctionality hypothesis, Shavitt and Nelson (1999) found that
people make fewer personality or individuating attributions for utilitarian products (e.g., aspirin).
In the study of domains we mentioned previously (Berger & Heath, 2005a), respondents rated
each domain on functionality. Supporting the afunctionality hypothesis, there was a high
negative correlation (r = -.90) between functionality and identity-relatedness; the less people saw
others’ choices as being determined by function, the more they used it to infer identity.
Our signaling logic implies that a particular cultural practice will act more effectively as a
signal when people could easily choose other cultural practices that have the same functional
value. A Burberry raincoat is a fashion statement even on a rainy day because other raincoats
could have served the same function at less expense. A black leather jacket with metal spikes
sends a strong signal because a plain brown one would have kept the wearer just as warm.
Another implication of our signaling logic is that individual tastes (e.g., fashion trends)
can identify themselves as identity-signaling by strategically violating functionality. Sunglasses
are more of a fashion statement indoors than out. A cap is more of a fashion statement when the
bill is turned backwards so it no longer serves its functional purpose. Afunctional behaviors are
also good signals. Beehive hairdos and mohawks are time-consuming to maintain, tattoos and
body-piercing are painful, high-heeled shoes are difficult to walk in, goldfish-swallowing and
phone-booth stuffing are obviously pointless, baggy Zoot suits waste cloth (indeed the War
Production Boards in WWII banned them because they wasted material). The afunctionality
Divergence 25
hypothesis helps us understand why Zoot suits are baggy, beehive hairdos high, and low-riders
low.
Summary: Where Do People Diverge?
Our identity-signaling approach answers the question of where divergence takes place
and why. Divergence occurs more in music tastes than toothbrushes because music tastes are
seen as signals of identity and people diverge to maintain the clarity of their signal. The social
coordination highlighted by our signal model ensures that individuals must express themselves in
the same domains others use to infer identity. Furthermore, it predicts that afunctional domains
are better suited for identity inferences because those domains make it easier to determine
meaning.
These results are difficult for prior approaches to explain. The uniqueness literature
argues people should diverge when they feel the most similar, but since people are more likely to
share the same pen or bike light than they are to have the same shirt, the uniqueness literature
would predict a pattern of divergence across domains that is opposite to the one found here.
Similarly, theories of novelty-seeking suggest that people should diverge when they are bored,
but without a specific theory of why boredom should vary by domain, it cannot explain domain
differences in divergence.
Our approach also improves on previous domain approaches suggesting when we should
see convergence versus divergence. In functional choices or those more likely to have a right
answer, social influence should elicit convergence. In a dot estimation task, for instance, others’
estimates will provide information that leads to convergence, as long as we think the quality of
their vision is similar to ours. Others’ type, however, should play a larger role in judgmental or
identity-oriented choices. People should converge with others of the same type, but diverge from
Divergence 26
different types; our friends using a catchphrase should prompt us to use it more, but other groups
using it may lead us to diverge.
So far our signaling model has been helpful in addressing some questions that have been
unresolved by prior approaches, but the current model still faces a serious challenge. If types
just want to be identified by others then why do many signals end up being costly or subtle?
Why would people get body-piercings or spend countless hours listening to music if they could
just wear t-shirts with smiley faces or motorcycle jackets? The probability of correct signal
identification presumably increases with observability, so why would people select anything but
the cheapest and most observable signals?
For simplicity, our first two models have assumed that everyone wants to be identified as
their own type. But this is too simple: the middle class might want to be treated as wealthy, the
unsophisticated as sophisticated, and the tragically unhip as hip. We now expand the model by
adding poseurs who want to be treated as members of a type they don’t belong to and we
investigate how their presence influences the signals people use and the way people diverge. We
investigate, in particular, a particularly stringent form of divergence--when people abandon
practices and tastes they once held.
Who Drives Divergence? Signal Poaching and Abandonment
It seemed only yesterday that Von Dutch trucker hats were worn by half the aggressively stylish people in the world. Now they are scorned in the hipster circles that only recently flaunted them (Lindgren, 2004, C4). So far we have discussed how signaling identity causes people to diverge in the tastes
they select, but there is also another type of divergence, namely taste abandonment. Not only do
different types of people select different tastes to maintain clear signals (e.g., teenagers choose
different music than their parents), but in some cases people also abandon tastes they previously
Divergence 27
held. Catchphrases die, people stop wearing certain styles of clothes, and fans abandon the very
musical artists they once followed zealously. Why?
We suggest individuals may abandon tastes when they are adopted by other social types
and thus their signal value is lost. Just as tastes gain meaning when they are associated with
particular types of individuals, they can lose meaning when they are adopted by other types.
Abandonment Hypothesis: individuals of one type will abandon previously held tastes when adoption by other types reduces their signaling value.
If Type A people ride Harley motorcycles, and Type B people (say brat pack Hollywood actors
or suburban accountants) start riding them, then an observer who sees someone riding a Harley
will find it more difficult to determine the person’s type. Consequently Type As may abandon
the taste because its signal value has been diluted/polluted.
In many cases, abandonment is driven by poseurs. In our basic model, types wanted to
be identified as their own type, but in many cases people of one type might prefer to be
considered another type. These poseurs may express another type’s signal in the hope of taking
on some of the attributes of that type. Teenagers may purchase fake Prada handbags to appear
rich and rich people may adopt teenage slang in an attempt to seem hip.
Model 3: Identity Poaching and Abandonment
To explore identity poaching and fashion abandonment, let us consider a sub-type, Type
BA. These poseurs are members of Type B, but prefer to be treated as members of Type A.
Formally, they want their interaction partner to choose action α, but other actors consider them
members of B, and prefer to choose action β. Furthermore, to allow the possibility of dynamic
change over time, consider a repeated version of the preceding game. Each period could be a
day, a week, etc., but in each period all actors choose to signal and then have the chance to
interact.
Divergence 28
Understanding the Model
Over time, poaching will undermine the signal, leading to abandonment. Consider a
domain (e.g., music tastes) where Type A has coordinated on signal i (e.g., hip-hop knowledge),
and Type B has coordinated on signal j (e.g., pop knowledge) and that signal costs are stable
over time. For the moment, take it as given that it is cheaper for Type A to signal hip-hop
knowledge than it is for Type BA. Members of Type A choose hip-hop and Type B choose pop,
and they treat others as members of their type when they display the appropriate musical taste.
Members of Type BA would prefer to learn hip-hop but most will find it too costly and will
continue with pop. However, some who do not find it too costly will switch to hip-hop.
If an actor sees someone with knowledge of pop, he can be sure they are of Type B. If an
actor sees someone with knowledge of hip-hop, the person is likely Type A, but there is some
small chance that the other is a poseur (BA). As long as this chance is small, Type A people will
continue to interact with hip-hop signalers, risking the chance of interacting with a poseur for the
greater reward of meeting one of their own.
Suppose, however, MTV starts playing more hip-hop, making it cheaper for poseurs to
learn about it (see Polhemus, 1994 for a discussion of media’s role in facilitating imitation).
Over time, more and more Type BA people will choose hip-hop, and the effectiveness of hip-hop
as a signal of identity will decline. If the number of poseurs who adopt hip-hop grows
sufficiently, the signaling will break down and the taste will be abandoned.
In addition to explaining why tastes are abandoned, the key insight of this model is that
the tastes that are most effective at signaling identity are those that maximize the cost differences
between groups, imposing a low cost for in-type members and a high cost for out-type members.
In the example above, poseurs were originally kept out by the high cost of knowing about hip-
Divergence 29
hop. But if the difference in costs were to disappear (because of MTV or some other factor),
poseurs enter, dilute the taste’s signal value, and signaling breaks down.
What Prior Literature Says About Abandonment
Sociologists agree that individuals make cultural choices to set themselves apart from
members of other social categories (Bourdieu, [1979] 1984; Davis, 1992; DiMaggio, 1982;
Dooley, 1930; Gans, 1974; Levine, 1988; Robinson, 1961; Simmel, 1904 [1957]; Veblen,
1899[1912]), but they have primarily focused on processes in which groups with traditionally
high status (e.g., wealthy, educated groups) diverge from traditionally low status adopters (see
Pesendorfer, 1995 and Bernheim, 1994 for economic treatments of the topic). Bourdieu ([1979]
1984) finds that, when asked what scene would make a great photo, people with higher levels of
education tend to skip over objects most likely to be popular with the masses (e.g., first
communion or sunset). Other work has shown that educated people tend to actively dislike
music associated with people of lower average education (e.g., rap, country and heavy metal,
Bryson, 1996) and that in the early 1900s, whites distanced themselves from the first names used
by blacks (London & Morgan, 1994).
Emphasizing that groups compete over identities in the broader social environment
provides a healthy corrective to views that presume individuals construct their own identities in
isolation, but unfortunately, the sociological literature has focused too narrowly on situations
where people compete over identities related to status as defined by particular macro-
demographic characteristics. This literature is based primarily on the trickle down theory of
fashion (Barber, 1957; Foley, 1893; Rae, 1834; Robinson, 1961; Simmel, 1904 [1957]; Veblen,
1899 [1912]), which suggests people adopt from those above them in the status food chain.
Fashions are initiated by the higher class and imitated by the lower classes. But, once the lower
Divergence 30
classes have adopted, the signal value is lost and consequently high class people abandon the
taste. In the context of social groups in the United States, this literature has essentially
endeavored to explain why Northeastern WASPs seek to distance themselves from every other
group in the country (e.g., blue collar workers, Southerners, Catholics, ethnic minorities).
As Mark (2003) notes, the original emphasis of the distinction literature is overly
asymmetric, focusing too much on high status people distancing themselves from low status
others. Many examples of abandoning tastes are not easy to classify in terms of a
unidimensional status ordering. Teenagers reject catchphrases once they creep into their parent’s
lexicon and, just as whites select names to distance themselves from blacks, blacks often
abandon clothing styles or slang that are adopted by whites. Fashion does not solely originate
among the upper classes (Davis, 1992), and, in fact, often comes from low-status or somewhat
marginalized groups. Much of fashion has been started “not so much by the upper or even
middle classes, as by the déclassé, anti-class youth, and counterculture. Long hair, head bands,
beads, tie-dyed apparel...not only mock the materialistic status symbols of the established
classes, but have successfully spread into the enemy camp, Fifth avenue and Main Street, where
they have caught on and been copied,” (Blumberg, 1971, p. 493).
While some theoretical approaches have touched on the notion of bidirectional influence
(e.g., mass market or trickle across theories--Sproles, 1981, 1985; King, 1963) and subcultural
leadership (e.g., Blumberg, 1974; Field, 1970), they have not systematically discussed how the
process works, and in particular they have not recognized that adoption by mainstream types will
lead the counterculture to abandon tastes.
Divergence 31
Our identity-signaling perspective takes a more general approach to divergence; rather
than divergence driven by status, we suggest divergence is driven by people’s desire for
beneficial interactions.
Experimental Evidence on Abandonment
Some experimental evidence has shown that taste abandonment is not driven purely by
macrodemographic status. Berger and Heath (2005a) examined whether adoption by other
students would lead students to abandon tastes. In the first stage of the experiment, students in a
mass testing session provided their preferences in a number of taste domains. In the car brand
domain, for instance, participants were presented with five car brands (Acura, BMW, Mercedes-
Benz, Volvo, Lexus) and asked to circle which they preferred. Two to three weeks later,
participants came to the lab in small groups for a seemingly unrelated study about how people
discuss their preferences. As they walked in, a research assistant was working at a side table in
the experiment room, visibly entering data. After an initial exercise, the experimenter
apologized, saying he had run out of copies for the second portion and had to go make more.
While the experimenter was gone, participants were asked to help the assistant tabulate some
preference data (to fill the time). Each participant was given a stack of surveys that had been
carefully tailored to suggest that in certain target domains, other respondents shared the
preferences they had expressed in the mass testing session (e.g., 8 out of 10 of the data sheets
they tallied showed choice of the same car brand that they chose previously). When the
experimenter returned with the copies, participants completed a preference questionnaire which
included the target domains. They then met as a group to discuss which options they had
selected.
Divergence 32
We hypothesized that when students learned their taste was shared by a majority of other
students, they would think that the taste was not a good signal of their type of student (e.g., jock,
geek, etc.). Indeed, results of a pre-test found that students who were told that 8 of 10 (vs. 2 of
10) other students shared one of their preferences were more likely to believe the preference was
held by people outside their own group. This dilution in signal value should lead people to
abandon tastes in identity-related domains.
Supporting the abandonment hypothesis, participants who were confronted with similar
responses by 6-8 others abandoned some of the tastes they had expressed in the earlier
questionnaire. Furthermore, supporting the domain hypothesis, they were more likely to
abandon tastes in domains that were afunctional and more identity relevant. A mean split on
identity inferences found that participants diverged 28% of the time in domains used to infer
identity (e.g., music preference) but only 17% of the time in domains not as likely to be used to
infer identity (e.g., bike lights). (This pattern of divergence was not echoed in a control
condition where people were not exposed to others’ preferences.) Because both the participants
and supposed adopters were students, one cannot explain the results in terms of status-based
divergence.
In another study, we tried to create a situation in which people could actually see that a
certain taste no longer had value in distinguishing between types. In the summer of 2004, the
Lance Armstrong Foundation started selling yellow Livestrong wristbands to support cancer
awareness and research. Originally, the wristbands were worn primarily by athletes, but they
later caught on more broadly and spread contagiously in the general public (see Walker, 2004).
During the upswing of this trend, we distributed these wristbands to various dorms on a
university campus to examine how signal dilution might affect abandonment. Specifically, we
Divergence 33
sold 38 wristbands to one dorm of 60 people (high prevalence condition), and half that many
were distributed to each of two other same-sized dorms on separate parts of campus (low
prevalence conditions). A week later, researchers returned to the dorms and, through an
unrelated questionnaire, measured whether dorm members were wearing the wristband they had
purchased. We predict people in the high prevalence dorm should be less likely to wear the
wristband because its meaning has become more diluted. Most dorm members weren’t athletes,
and consequently, the more people in the dorm wearing the wristband, the greater the percentage
of wristband wearers a person interacts with who are not athletes (and are everything from
computer scientists to hardcore partiers). This should dilute the taste’s signal value. Indeed,
abandonment was greater in the high prevalence dorm, where a smaller percentage of members
actually wore the wristband to house meeting (58% vs. 88% in low-prevalence dorms).
These experimental results are difficult for the uniqueness literature to explain. A greater
number of others sharing one’s car preference or possessing the same wristband might have
made people feel slightly less unique, but the differences used here (e.g., 38 vs. 19 others owning
a wristband) are nowhere near the extreme similarity manipulations used in the traditional
uniqueness experiments (e.g., 80% similarity to 10,000 students on 30 attitude items, Ganster,
McCuddy & Fromkin, 1977).
Abandonment Based on Who is Adopting
We can also use the framework of model three to think about signaling in a richer setting
populated by multiple types of people. Overall, adoption by other types should lead to
divergence, but we predict that divergence may be greater when the adopting type is more
Divergence 34
dissimilar to the current taste holders.1 Presumably, larger type differences imply that people
experience higher costs when they are misidentified as members of the wrong type. That is, the
differences between correct and incorrect identification become larger when the types are more
dissimilar. Snowboarders might not want to be identified as skiers, but they might prefer that
mistake to being misidentified as a golfer.
Alternatively, if tastes diffuse across similar types, then tastes that have been adopted by
very dissimilar types are likely to have been adopted by more people (all the types in between),
thus the signal has experienced a higher degree of dilution. Consequently, one would expect that
divergence would increase with dissimilarity.
To test this notion, Stanford undergraduates were told to imagine their groups of friends
engaged in a certain behavior that had been adopted by some other group (Berger & Heath,
2005b). For each of 15 social groups (e.g., 40 year old white business executives, janitors,
college professors, suburban teenagers, etc.) participants were asked to rate how many people
(outside that group) were likely engaging in the behavior, given that a given social group had
adopted it. Another set of raters rated the groups on similarity. Supporting our prediction,
adoption estimates were highly negatively correlated with similarity (r = -.74); people assumed
more people had adopted the taste if the adopting group was dissimilar.
It is worth noting that there exists a countervailing effect that should lead to relatively
increased divergence when adopting types are extremely similar. The previous two points were
based on the costs of misidentification, but the probability of misidentification should also vary.
When types are more similar, they are more likely to overlap in their tastes and they may be
1 We use similarity in a broader sense than mere demographic similarity. A hip teenager may be demographically dissimilar to a 50-year-old business executive, but if the executive has a shaved head, wears black, and makes independent movies, the two may be considered similarly hip even though they differ demographically.
Divergence 35
more likely to be confused with each other. Cooper and Jones (1966) found that people diverged
from an obnoxious lab partner, but more so when the partner seemed similar (e.g., father had a
similar occupation). Also, similar types should be more likely to encounter each other than to
encounter more distant types.
Combining these factors, we predict a curvilinear relationship between similarity of the
adopting type and divergence; people will be more likely to abandon a practice when it is
adopted by dissimilar types, but they may also increase abandonment if it is adopted by types
who are extremely similar.
Similarity Hypothesis: the influence of similarity of the adopting type on divergence will be curvilinear; both extreme dissimilarity and extreme similarity will lead to increased divergence. We tested this hypothesis (as well as the effect of adopter status) in a study regarding
catchphrase usage (Berger & Heath, 2005b). Stanford undergraduates were told to imagine their
group of friends liked to say a particular catchphrase that no one else was saying. Then they
were asked to rate how their use of the phrase would change if it were adopted by various other
social groups (the same 15 used in the previous study). Other sets of students rated each group
on similarity, liking, or demographic status.
For 14 of the 15 groups, participants said they would be likely to abandon the catchphrase
if others adopted it (see Figure 2). Abandonment was most pronounced for highly dissimilar
groups but was also noticeable for similar groups. The curvilinear effect of similarity held even
when we controlled for liking and demographic status.
These results are consistent with the identity-signaling approach but inconsistent with
other theories. Theories of conformity obviously predict convergence when other people adopt a
taste, not divergence. Other theories might predict that people diverge from disliked others (e.g.,
Divergence 36
Bryson, 1996; Wood, Pool, Leck, & Purvis, 1996), but the similarity result holds even when
liking is controlled. This makes intuitive sense-- kids may like their parents a lot, but diverge
when parents adopt their slang or way of dress. Sociological models of status also have
difficulty explaining these results: the status of the adopting group was not significantly
correlated with divergence and people diverged even when their taste was adopted by a higher
status group.
An additional study tested both the similarity and domain hypotheses, examining how
similarity of the adopting group and the particular taste domain would influence divergence.
Stanford students were given 10 domains (varying in identity-relatedness) and asked to imagine
that members of a certain social group had started to copy their preference in that domain (e.g.,
Princeton students had started adopting their favorite type of music). They then indicated what
level of adoption by the other group would lead them to abandon the taste (e.g., would they
abandon if 10% of Princeton students had adopted? 50%? Never?). The other group that
adopted was varied in a between subjects design; we chose three groups (40 year old business
executives, inner city teens, and Princeton students) that were rated equivalently on a measure of
liking but varied on similarity.
Both divergence predictions were supported: people were more likely to diverge in
identity related domains (e.g., clothing versus bike lights), and there was a curvilinear
relationship between amount of divergence and the similarity of the adopting group--participants
diverged more when the adopting group was either extremely dissimilar (i.e. 40 year old business
executives) or extremely similar (i.e. Princeton students) than when it was in the middle (inner
city teens); see Figure 3. Furthermore, consistent with our approach, there was also an adopter
similarity x domain interaction: both similar and dissimilar groups prompted divergence in
Divergence 37
domains used to infer identity, but similarity had no effect in domains where tastes are not used
to infer identity (e.g., toothpaste). This interaction provides particularly clear support for
signaling and not mere association or liking--people care about the clarity of their signals, but
only in identity related domains.
Summary: Who drives divergence and why?
Our identity-signaling approach answers the question of who drives divergence and why.
People diverge when other types adopt their practices and their identity signals become diluted.
This relationship is curvilinear--people diverge from similar others (when the likelihood of
misidentification is high), and they diverge more from very dissimilar others (when the cost of
misidentification is high).
A major advantage of our model is the capacity to explain why divergence occurs even
when the adopting group is dissimilar. Previous literature has emphasized similarity-- the
uniqueness literature argues divergence is driven by similar others and social identity research
has argued groups feel increased pressure to differentiate themselves from similar outgroups
(Brown & Abrams, 1986; Diehl, 1988; Roccas & Schwartz, 1993). But previous research has
not explained why groups might want to differentiate themselves from dissimilar others. This
seems to neglect some key phenomena in the environment: punks or inner city teens would
probably abandon tastes that were adopted by suburban teens (or, even more serious, suburban
parents).
Although we have argued that pressures to diverge are more general than suggested by
literature that has focused on macro-demographic status, that literature has correctly recognized
that some social groups may find the problems of poaching to be particularly acute. If people
like to appear wealthy, the rich must keep switching signals. But our model goes beyond the
Divergence 38
traditional sociological emphasis on macro-status because it predicts the divergence race will
also be more intense for types that are highly imitated because they are interesting or cool.
When suburban kids imitate inner city teens or when straight metrosexuals imitate the signals of
gay men, the original groups will also be forced to change signals in order to maintain signal
clarity.
Our research also provides some experimental backing to the notion that tastes may die
when they become too popular (the classic example of a non-conformity effect, e.g., Lieberson,
2000). “Smiley-face t-shirts were cast off as uncool and the word ‘acid’ was dropped from club
names and music genre classifications as soon as ‘acid house’ became a term familiar to general
readers of national newspapers,” (Thorton, 1997, p 6.). This presents an interesting macro-level
paradox that points out one limit to conformity processes. While people may be drawn to
popular things because others like them, if too many people hold the same taste it loses both its
value as a unique attribute and as an identity signal, and people will abandon it.
Costly Signals
As discussed previously, signals are often costly, yet the prediction of Model 1 was that
signals would be visible and low cost. The idea of poaching provides insight into why costly
signals may persist. While people might prefer to signal with the least possible cost, cheap
signals are the most likely ones to be poached. Thus costly tastes should be more likely to
persist over time because they impose barriers that make adoption relatively difficult: Goth style
“is by its very nature off-putting and therefore has avoided being drawn into the mainstream”
(Polhemus, 1994, p. 97). Tastes that are painful (e.g., piercing), require lots of money (e.g.,
expensive jewelry), or take a long time to acquire (e.g., a love of chess) make poaching harder,
so they may resist signal dilution. Knowledge based tastes (e.g., an appreciation of Gregorian
Divergence 39
chant) may be particularly resistant to poaching because they cannot just be purchased by anyone
willing to put up the money. In many situations costly signals may evolve in a relatively
unplanned way (e.g., certain tastes happen to be more costly, and thus happen to persist because
they are less likely to be poached by others). But in other cases, types may deliberately select
costly tastes to foil poachers: the jazz musicians who developed the New York sound
“deliberately sought to restrict white identification by producing a jazz which was difficult to
listen to and even more difficult to imitate” (Hebdige, 1987, p. 47).
Visibility
Visible signals create a tension: visibility makes signals easier to see but also easier to
poach. If Harley Davidson riders wear black leather jackets, it will be easier to associate jacket-
wearing with Harley-riding than with some other, more private, signal (e.g., a tattoo that is
typically hidden). But this visibility also makes it easier for poseurs to recognize the signal and
copy it. Visible signals should lose meaning faster, and consequently, we should see faster
turnover in taste domains that are more public.
Visibility also interacts with opportunity costs. Some highly visibility signals may be
less likely to be copied because they impede interactions with other types (e.g., Mohawks may
make it difficult to find a job). Conversely, some types may select a low visibility signal to
allow for easy interactions with other types but still identify each other. Rubinstein (1995) notes
that before the 1980s, most gay men felt pressure to remain invisible to mainstream culture.
Consequently, they had to come up with subtle signals that could be identified by other gay men
but not by outsiders. To do so, they adopted what has been termed the Old Clone look, wearing
“jeans, lumber shirts, jackets, and heavy boots and sporting a mustache and
sunglasses…instantly recognizable by other gay men, the look would not offend at work, for
Divergence 40
most non-gay colleagues would miss its significance” (p. 215). By selecting subtle signals, they
were able to signal only to others of the same type.
General Discussion
We have proposed an identity signaling approach to divergence: people diverge to signal
their identity to others so that they can enjoy more fulfilling social interactions. Importantly, this
process is socially driven. Instead of focusing on idiosyncratic individual drives, we focus on
how people communicate and receive meaning. Tastes acquire meaning when they are
expressed by similar types of people and they lose meaning when they are adopted by other
types.
Our approach answers a number of questions that have been unanswered by previous
approaches to divergence. Divergence does not occur idiosyncratically by individuals, rather
types of individuals diverge when signals of identity have been undermined. Divergence should
be more prevalent in domains used to infer identity and these domains are likely to be ones in
which choice is not based on function. Divergence is not driven by similar others, but occurs
whenever adoption by other social types reduce a taste’s signal value.
Although the game theoretic model we use invokes the language of choice, we do not
believe that divergence is necessarily conscious or intentional. Social influence can often occur
nonconsciously (Pronin, Berger, & Molouki, 2005), and research on mere exposure (Kuntz-
Wilson & Zajonc, 1980) and social mimicry (Chartrand & Bargh, 1999) suggests people may
come to like things others like or to imitate their behavior, all without awareness that they have
been influenced. (For example speech patterns and pronunciations are often quite subtle and
may be hard to acquire consciously but they demonstrate some of the divergence processes we
have described, Eckert, 1989, 2000). Similarly, people may diverge from others because others’
Divergence 41
tastes just don’t seem “right” for them (see work on prototype matching, Niedenthal, Cantor, &
Kihlstrom, 1985). Yet the formal model is useful because we assume that, whether selected
consciously or unconsciously, some patterns of practices are more likely to lead to an
equilibrium that allows for efficient coordination, and those practices are more likely to persist.
Divergence and Culture
Cultural psychologists suggest that relative to Americans, members of East Asian
cultures prefer less uniqueness (Aaker & Schmitt, 2001; Kim & Markus, 1999; Kim & Drolet,
2003). As a result, uniqueness based explanations have trouble explaining why we still see much
divergence in East Asian cultures; just as the U.S has skaters, surfers, punks, etc. Japan has
kawaii (baby doll), yamamba (mountain witch), bodikon (skimpy skirts and high-heeled boots),
(Mead, 2002). A signaling approach, however, does not have such difficulty. Even though the
self is less focal (e.g., Morris & Peng, 1994), members of East Asian cultures should diverge for
the same reason Americans do, to signal clear identities as members of types or collectives.
Indeed, if collectives are more important in East Asian cultures, then identity signaling dynamics
may be even more important (Wong & Ahuvia, 1998).
The Dynamic Meaning System
The fact that meaning changes socially suggests that even if individuals want to express a
consistent identity, they may have to adopt different tastes over time. A rich white male may
have originally worn Tommy Hilfiger to signal his upscale preppy identity, but the adoption of
Hilfiger clothing by hip-hop artists in the early 1990’s meant that those clothes came to signal an
entirely different identity. Consequently, rich white men may have needed to diverge just to
maintain a consistent signal.
Divergence 42
While our discussion has focused on active divergence when signals lose their value, the
identity signaling suggests convergence and divergence happen even in the absence of these
drives. People may end up conforming not because they want to imitate others, but because they
do not want to send out the wrong signals. When hairstyles were trending longer in the 1960s,
the typical person may have been forced to track the expanding length because they did not want
to send a strong signal on either end--too short and they are considered stodgy and conservative,
too long and they are considered to be hippies. The social nature of signaling means that even
when individuals want to avoid sending a particular signal, they may need to alter their behavior
to stay in step with the social dynamic.
Cycles of Fads and Fashions
Seeing tastes as signals of identity also helps understand why fads and fashions fluctuate.
Fads are not limited to frivolous consumables (e.g., Beanie Babies and Hula Hoops, see Aguirre,
Quarantelli, & Mendoza, 1988; Arthur & Lane, 1993; Blumer, 1969; Fearon, 2005; Robinson,
1958). Indeed, fad-like processes of rapid adoption and abandonment occur in everything from
management practices (e.g., Total Quality Management) to public policies (e.g., whole math) to
academic theories (e.g., consider Strauss and Quinn’s 1997, p. 12, comments about how theories
rise and fall in anthropology). While lay theories of fads and fashion suggest people abandon
practices because they just “get tired of them” (some disciplinary theories suggest the same, see
Sproles, 1981) a signaling perspective provides more insight. Whenever tastes signal identity,
and especially when they are visible and thus easy for poseurs to poach, we should see upswings
in popularity followed by abandonment as taste holders diverge because the taste’s signaling
value has been diluted.
Identity and Social Influence
Divergence 43
More attention needs to be given to the role of identity in social influence. Much social
influence research suggests that people will converge to the behavior of others, but the evidence
provided here suggests that when the others are members of a different social type, people may
in fact diverge. These results are difficult to explain using the traditional distinction between
normative and information social influence (e.g., Deutsch & Gerard, 1955). The studies
contained few vehicles for rewards and punishments and, if anything, the information provided
by others’ choices should have led people to converge. The referent informational influence
approach (e.g., Hogg & Turner, 1987; Turner, 1982) does a better job of taking the identity of the
influencer into account and deserves more attention, but this theory focuses predominantly on
conformity and thus has less to say about divergence. Consequently, more work needs to be
done to incorporate identity into social influence. People’s tastes may be greatly influenced by
the tastes and behavior of others, but teens and their parents, jocks and geeks, and residents of
Shanghai and New York know that social influence may lead people to diverge as well as
conform.
Divergence 44
References
Aaker, J. (1997). Dimensions of brand personality. Journal of Marketing Research, 34, 342-352.
Aaker, J., & Schmitt, B. (2001). Culture-dependent assimilation and differentiation of the self.
Journal of Cross Cultural Psychology, 32, 561-576.
Abelson, R. P., & Prentice, D. A. (1989). Beliefs as possessions -- A functional perspective. In
A. R. Pratkanis, S. J. Breckler, & A. G. Greenwald (Eds.), Attitude structure and function
(pp. 361-381). Hillsdale, NJ: Erlbaum.
Acker, M., & McReynolds, P. (1967). The need for novelty: a comparison of six instruments.
Psychological Record, 17, 177-82.
Aguirre, B.E., E.L. Quarantelli & Mendoza, J.L. (1988). The collective behavior of fads: The
characteristics, effects and career of streaking. American Sociological Review, 53, 569-584.
Allen, V.L. (1965). Situational factors in conformity. In L. Berkowitz (Ed.), Advances in
experimental social psychology, 2. New York: Academic, 133-170.
Allport, F. H. (1924). Social psychology. Boston: Houghton Mifflin.
Ariely, D., & Levav, J. (2000). Sequential choice in group settings: Taking the road less traveled
and less enjoyed. Journal of Consumer Research, 27, 279-290.
Arthur, W. B., & Lane, D.A. (1993). Information contagion. Structural Changes and Economic
Dynamics, 4, 81-103.
Asch, S.E. (1951). Effects of group pressure upon the modifications and distortion of judgments.
In H. Guetzkow (Ed.), Groups, leadership, and men. Pittsburgh, PA: Carnegie Press.
Ball, D., & Tasaki, L. (1992). The role and measurement of attachment in consumer behavior,
Journal of Consumer Psychology, 1, 155-172.
Divergence 45
Banerjee, A.V. (1992). A simple model of herd behavior. Quarterly Journal of Economics, 107,
797-817.
Barber, B. (1957). Social Stratification. New York: Harcourt, Brace, and Company.
Barker, O. (2004). Everything is so 5 minutes ago. USA Today, January 26, 1D.
Bass, F. M (1969). A new product growth model for consumer durables. Management Science,
13, 215-227.
Bearden, W. O., & Etzel, M.J. (1982). Reference group influence on product and brand purchase
decisions. Journal of Consumer Research, 9, 183-194.
Belk, Russell W. (1981). Determinants of consumption cue utilization in impression formation:
An associational deviation and experimental verification, In K. Monroe (Ed.), Advances in
Consumer Research, 8, Ann Arbor, MI: Association for Consumer Research.
Belk, R.W. (1988). Possessions and the extended self. Journal of Consumer Research, 15, 139-
167.
Belk, R.W., Bahn, K.D., & Mayer, R.N. (1982). Developmental recognition of consumption
symbolism. Journal of Consumer Research, 9, 4-17.
Berger, J. A., & Heath, C. (2005a). Where do people diverge from others? Tastes as signals of
identity. Under Review.
Berger, J.A., & Heath, C. (2005b). Why Fads and Fashions Fade: Cultural Tastes as Signals of
Identity. Working paper.
Bernheim, B.D. (1994). A theory of conformity. Journal of Political Economy, 102, 841-77.
Bikhchandani, W. Hirshleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom, and
cultural change as informational cascades. Journal of Political Economy, 100, 992-1026.
Divergence 46
Blumberg, P. (1974). The decline and fall of the status symbol: Some thoughts on status in post-
industrial society. Social Problems, 21, 480-498.
Blumer, H. (1969). Symbolic Interactionism: Perspective and Method. Englewood Cliffs, NJ:
Prentice-Hall.
Bond, R.A., & Smith, P.B. (1996). Culture and conformity: A meta-analysis of studies using
Asch's (1952, 1956) line judgment task. Psychological Bulletin, 119, 111-137.
Bourdieu, P. ([1979] 1984). Distinction: A Social Critique of the Judgment of Taste, translated
by Richard Nice. Cambridge, Mass: Harvard University Press.
Brewer, M. (1991). The social self: On being the same and different at the same time.
Personality and Social Psychology Bulletin, 17, 475-482.
Brewer, M. B. (1993). The role of distinctiveness in social identity and group behavior. In M. A.
Hogg & D. Abrams (Eds.), Group motivation: Social psychological perspectives (pp. 1-16).
Hertfordshire, UK: Harvester Wheatsheaf.
Brewer, M.B., & Pickett, C.L. (1999). Distinctiveness motives as a source of the social self. In
Tyler, T. R., & Kramer, R.M. (Eds.), The psychology of the social self (pp.71-87). Mahwah,
NJ: Lawrence Erlbaum Associates.
Brewer, M.B., & Pickett, C.L. (2002). The social self and group identification: Inclusion and
distinctiveness motives in interpersonal and collective identities. In J. Forgas & K. Williams
(Eds.), The social self: Cognitive, interpersonal, and intergroup perspectives (pp. 255-271).
Philadelphia: Psychology Press
Brewer, M.B., & Roccas, S. (2001). Individual values, social identity, and optimal
distinctiveness. In C. Sedikides and M. Brewer (Eds.), Individual Self, Relational Self,
Collective Self (pp. 219-237). Philadelphia: Psychology Press.
Divergence 47
Brown, R. J., & Abrams, D. (1986). The effects of intergroup similarity and goal
interdependence on intergroup attitudes and task performance. Journal of Experimental
Social Psychology, 22, 78-92.
Bryson, B. (1996). Anything but heavy metal: Symbolic exclusion and musical dislikes.
American Sociological Review, 61, 884-899.
Cattell, R.B. (1975). Personality and motivation: Structure and Measurement. New York:
Harcourt, Brace, & World, Inc.
Campbell, J.D. (1986). Similarity and uniqueness: The effects of attribute type, relevance, and
individual differences in self-esteem and depression. Journal of Personality and Social
Psychology, 50, 281-294.
Chartrand, T.L., & Bargh, J.A. (1999). The chameleon effect: The perception-behavior link and
social interaction. Journal of Personality and Social Psychology, 76, 893-910.
Cooper, J., & Jones, E.E. (1969). Opinion divergence as a strategy to avoid being miscast.
Journal of Personality and Social Psychology, l3, 23-30.
Davis, F. (1992). Fashion, Culture and Identity. Chicago: Chicago University Press.
Deutsch, M., & Gerard, H.B. (1955). A study of normative and informational social influences
upon individual judgment. Journal of Abnormal and Social Psychology, 51, 629-636.
Diehl, M. (1988). Social Identity and minimal groups: The effects of interpersonal and
intergroup attitudinal similarity on intergroup discrimination. British Journal of Social
Psychology, 27, 289-300.
Divergence 48
DiMaggio, P. (1982). Cultural capital and school success: The impact of status culture
participation on the grades of U.S. high school students. American Sociological Review, 47,
189-201.
DiMaggio, P. (1987). Classification in Art. American Sociological Review, 52, 440-455.
DiMaggio, P., & Powell, W. (1983). The iron cage revised: Institutional isomorphism and
collective rationality in organizational fields. American Sociological Review, 48, 147-160.
Dooley, W.H. (1930). Clothing and Style. New York: D.C. Heath and Company.
Douglas, M., & Isherwood, B. (1978). The World of Goods: Towards an Anthropology of
Consumption. New York: W.W. Norton.
Duval, S. (1972) Conformity on a visual task as a function of personal novelty on attitudinal
dimensions and being reminded of the object status of the self. Unpublished doctoral
dissertation, University of Texas.
Eckert, P. (1989). Jocks and Burnouts: Social Identity in the High School. New York: Teachers
College Press.
Eckert, P. (2000). Linguistic variation as social practice. Oxford, Blackwell.
Epley, N., & Gilovich, T. (1999). Just going along: Nonconscious priming and conformity to
social pressure. Journal of Experimental Social Psychology, 35, 578-89.
Escalas, J.E., & Bettman, J.R. (2003). You are what they eat: The influence of reference groups
on consumers’ connections to brands. Journal of Consumer Psychology, 13, 339-348.
Farley, F., & Farley, S. (1967). Extroversion and stimulus-seeking motivation. Journal of
Consulting Psychology, 31, 215-216.
Divergence 49
Fearon, J. (2005). Chunky shoes: A model of style choice and fashion trends with implications
for understanding “image” advertising. Working Paper.
Festinger, L. (1950). Informal social communication. Psychological Review, 57, 271-282.
Field, G. A. (1970). The status float phenomenon - The upward diffusion of innovation. Business
Horizons, 8, 45-52.
Foley, C.R. (1893). Fashion. Economic Journal, 3, 458.
Fromkin, H.L. (1968). Affective and valuational consequences of self-perceived uniqueness
deprivation. Unpublished Doctoral Dissertation, Ohio State University.
Fromkin, H.L. (1970). Effects of experimentally aroused feelings of indistinctiveness upon
valuation of scarce and novel experiences. Journal of Personality and Social Psychology, 16,
521-529.
Fromkin, H.L. (1972). Feelings of interpersonal undistinctiveness: An unpleasant affective state.
Journal of Experimental Research in Personality, 6, 178-182.
Gans, H.J. (1974). Popular Culture and High Culture. New York: Basic Books.
Ganster, D., McCuddy, M., & Fromkin, H.L. (1977). Similarity and undistinctiveness as
determinants of favorable and unfavorable changes in self-esteem. Paper presented at the
Midwestern Psychological Association, Chicago.
Goffman, E. (1951). Symbols of class status. British Journal of Sociology, 2, 294-304.
Graham, D. (1962). Experimental studies of social influence in simple judgment situations.
Journal of Social Psychology, 56, 245-26
Granovetter, R. M., & R. Soong (1986). Threshold models of interpersonal effects in consumer
demand. Journal of Economic Behavior and Organization, 7, 83-99.
Divergence 50
Gross, R., & Osterman, P. (1971). Individualism: Man in Modern Society. New York: Laurel.
Hebdige, D. (1987) Subculture: The Meaning of Style, London: Routledge.
Herek, G.M. (1986). The instrumentality of attitudes: Toward a neofunctional theory. Journal of
Social Issues, 42, 99-114.
Hogg, M.A., & Turner, J.C. (1987). Social identity and conformity: a theory of referent
informational influence. In Doise, W., & Moscovici, S. (Eds.), Current Issues in European
Social Psychology, Vol. 2 (pp. 139-182). Cambridge Univ. Press, New York.
Holman, R.H. (1981). Product as communication: A fresh appraisal of a venerable topic. In B.M.
Eris and K.J. Boering (Eds.), Review of Marketing (pp. 106-119). Chicago: American
Marketing Association.
Holt, D.B. (1997). Poststructuralist lifestyle analysis: Conceptualizing the social patterning of
consumption in postmodernity. Journal of Consumer Research, 23, 326-350.
Ianneccone, L.R. (1989). Bandwagons and the threat of chaos: Interpersonal effects revisited.
Journal of Economic Behavior and Organization, 11, 431-442.
Jetten, J., Branscombe, N.R., Schmitt, M.T., & Spears, R. (2001). Rebels with a cause: Group
identification as a response to perceived discrimination from the mainstream. Personality and
Social Psychology Bulletin, 27, 1204-1213.
Kahn, B.E. (1995). Consumer variety-seeking among goods and services. Journal of Retailing
and Consumer Services, 2, 139-148.
Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24,
163-204.
Kelley, H.H. (1973). The process of causal attribution. American Psychologist, 28, 107-128.
Divergence 51
Kernis, M.H. (1984). Need for uniqueness, self-schemas, and thoughts as moderators of the
false-consensus effect. Journal of Experimental Social Psychology, 20, 350-362.
Kim, H., & Markus, H.R. (1999). Deviance or uniqueness, harmony or conformity? A cultural
analysis. Journal of Personality and Social Psychology, 77, 785-800.
Kim, H., & Drolet, A. (2003). Choice and self-expression: the effect of cultural assumptions of
choice and uniqueness on variety-seeking. Journal of Personality and Social Psychology, 85,
373-382.
King, C. W. (1963). Fashion adoption: A rebuttal to the ‘Trickle-down’ theory. In S.A. Greyser
(Ed.). Toward Scientific Marketing (pp. 108-125). Chicago: American Marketing
Association.
Kleine, R.E., III, Kleine, S.S., & Kernan, J.B. (1993). Mundane consumption and the self: A
social identity perspective, Journal of Consumer Psychology, 2, 209-235.
Kuntz-Wilson, W., & Zajonc, R.B. (1980). Affective discrimination of stimuli that cannot be
recognized. Science, 207, 557-558.
Latane, B. (1981). The psychology of social impact. American Psychologist, 36, 343-356.
Leibenstein, H. (1950). Bandwagon, snob, and Veblen effects in the theory of consumers
demand. The Quarterly Journal of Economics, 65, 183-207.
Leibenstein, H. (1976) Beyond Economic Man: A New Foundation for Microeconomics.
Cambridge, MA: Harvard University Press.
Levine, L.W. (1988). Highbrow/Lowbrow: The emergence of cultural hierarchy in America.
Cambridge, MA: Harvard University Press.
Levy, S. J. (1959). Symbols for sale. Harvard Business Review, 33, 117-124.
Divergence 52
Lieberson, S. (2000). Matter of taste: How names, fashions, and culture change. Yale
University Press: New Haven.
Lindgren, H. (2004). Tearing up the runway in $400 cleats. New York Times. February 8, C4.
London, A.S., & Morgan, S. P. (1994). Racial differences in first names in 1910. Journal of
Family History, 19, 261-284.
Lynn, M., & Harris, J. (1977). Individual differences in the pursuit of self-uniqueness through
consumption. Journal of Applied Social Psychology, 27, 1861-1883.
Strang, D. & Macy, M. W. (2001). In search of excellence: Fads, success stories, and adaptive
emulation. American Journal of Sociology, 107, 147-182.
Mark, N. (2003). Culture and competition: Homophily and distancing explanations for cultural
niches. American Journal of Sociology, 68, 319-346.
McAlister, L. (1982). A dynamic attribute satiation model of variety-seeking behavior. Journal
of Consumer Research, 9, 141-150.
McCracken, G. (1986). Culture and consumption: A theoretical account of the structure and
movement of the cultural meaning of consumer goods. Journal of Consumer Research, 13,
71-84.
Mead, R. (2002). Shopping rebellion: What the kids want. The New Yorker, March 18, 104-111.
Mead, G.H. (1934). Mind, Self and Society. Chicago: University of Chicago Press.
Meyer, J., & Rowan, B. (1977). Institutionalized organizations: Formal structure as myth and
ceremony. American Journal of Sociology, 83, 340-363.
Milgrim, S., Bickman, L., and Berkowitz, L. (1969). Note on the drawing power of crowds of
different size. Journal of Personality and Social Psychology, 13, 79-82.
Divergence 53
Miller, C. M., McIntyre, S.H., & Mantrala, M.K. (1993). Toward formalizing fashion theory.
Journal of Marketing Research, 30, 142-157.
Moore, H. T. (1921). The comparative influence of majority and expert opinion. American
Journal of Psychology, 32, 16-20.
Morris, M.W., & Larrick, R, (1995). When one cause casts doubt on another: A normative
analysis of discounting in causal attribution. Psychological Review, 102, 331-355.
Morris, M.W., & Peng, K. (1994). Culture and cause: American and Chinese attributions for
social and physical events. Journal of Personality and Social Psychology, 67, 949-971.
Niedenthal, P. M., Cantor, N., & Kihlstrom, J. F. (1985). Prototype matching: A strategy for
social decision making. Journal of Personality and Social Psychology, 48, 575–584.
Park, C.W., Jaworski, B.J., & MacInnis, D.J. (1986). Strategic brand concept-image
management. Journal of Marketing, 50, 135-145.
Pesendorfer, W. (1995). Design innovation and fashion cycles. American Economic Review, 85,
771-792.
Peterson, R.A. (1978). The production of cultural change: The case of contemporary country
music. Social Research, 45, 292-314.
Polhemus, T. (1994). Street Style. United Kingdom: Thames and Hudson.
Prentice, D. A. (1987). Psychological correspondence of possessions, attitudes, and values.
Journal of Personality and Social Psychology, 53, 993-1003.
Pronin, E., Berger, J.A., & Molouki, S. (2005). Alone in a Crowd of Sheep: Asymmetric
Perceptions of Conformity and Their Roots in an Introspection Illusion. Under Review.
Rae, J. (1834). The Sociological Theory of Capital. London: Macmillan Company
Divergence 54
Ratner, R. K., & Kahn, B. E. (2002). The impact of private vs. public consumption on variety
seeking behavior. Journal of Consumer Research, 29, 246-258.
Richins, M. L. (1994a). Valuing things: The public and private meaning of possessions. Journal
of Consumer Research, 21, 504-521.
Richins, M.L. (1994b). Special possessions and the expression of material values. Journal of
Consumer Research, 21, 522-533
Robinson, D. E. (1961). The economics of fashion demand. Quarterly Journal of Economics, 75,
376-398.
Roccas, S., & Schwartz, S.H. (1993). Effects of intergroup similarity on intergroup relations.
European Journal of Social Psychology, 23, 581-595.
Rogers, E.M. (1983). Diffusion of Innovations. New York: The Free Press.
Rubinstein, R.B. (1995) Dress Codes. Boulder, Co: Westview Press.
Scott, W.R. (1995). Institutions and Organizations. Thousand Oaks, CA: Sage.
Schlenker, B. (1980). Impression Management. Belmont, CA: Wadsworth.
Shavitt, S. (1990). The role of attitude objects in attitude functions. Journal of Experimental
Social Psychology, 26, 124-148.
Shavitt, S., & Nelson, M. R. (1999). The social identity function in person perception:
Communicated meanings of product preferences. In G. R. Maio & J. M. Olson (Eds.), Why
we evaluate: Function of attitudes (pp. 37–57). Mahwah, NJ: Erlbaum.
Sherif, M. (1936). The Psychology of Social Norms. New York: Harper.
Simmell, G. ([1904] 1957). Fashion. American Journal of Sociology, 62, 541-548.
Divergence 55
Sirgy, M. J. (1982). Self-concept in consumer behavior: A critical review. Journal of Consumer
Research, 9, 287-300.
Smith, C.S. (1975). Self-definition change as a function of interpersonal similarity: A test of a
psychological spacing model. Unpublished master’s thesis, University of Kansas.
Smith, M. B. (1947). The personal setting of public opinions: A study of attitudes toward Russia,
Public Opinion Quarterly, 11, 507-523.
Smith, M. B., Bruner, J. S., & White, R. W. (1956). Opinions and personality. New York: Wiley.
Snyder, C.R. (1992). Product scarcity by need for uniqueness interaction: A consumer catch-22
carousel? Basic and Applied Social Psychology, 13, 9-24.
Snyder, C.R., & Batson, C.D. (1974). The balanced interpersonal perception of differences and
similarities: A model of psychological distance. Paper presented at the Western
Psychological Association, San Francisco.
Snyder, C.R., & Endelman, J.R. (1979). Effects of degree of interpersonal similarity on physical
distance and self-reported attraction: a comparison of uniqueness and reinforcement theory
predictions. Journal of Personality, 47, 492-505.
Snyder, C.R., & Fromkin, H.L. (1977). Abnormality as a positive characteristic: The
development and validation of a scale measuring need for uniqueness. Journal of Abnormal
Psychology, 86, 518-527.
Snyder, C.R., & Fromkin, H.L. (1980). Uniqueness, New York: Plenium.
Solomon, M. R. (1983). The Role of Products as Social Stimuli: A Symbolic interactionism
perspective. Journal of Consumer Research, 10, 319-329.
Divergence 56
Sproles, G.B. (1981). Analyzing fashion life cycles: Principles and perspectives. Journal of
Marketing, 45, 116–124.
Sproles, G. B. (1985). Behavioral science theories of fashion. In M. R. Solomon (Ed.). The
Psychology of Fashion (pp. 55-70). Lexington: D. C. Heath/Lexington Books.
Stafford, J.E. (1966). Effects of group influence on consumer brand preferences. Journal of
Marketing Research, 3, 68-75.
Strauss, C., & Quinn, N. (1997). A cognitive theory of cultural meaning. Cambridge, UK:
Cambridge University Press.
Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S.
Worchel & W. G. Austin (Eds.), Psychology of intergroup relations (pp. 7–24). Chicago:
Nelson-Hall.
Thornton, S (1995). Club Cultures: Music, Media and Subcultural Capital. Cambridge and
Oxford: Polity Press
Tian, K.T., Bearden, W.O., Hunter, G.L. (2001). Consumers’ need for uniqueness: Scale
development and validation. Journal of Consumer Research, 28, 50-66.
Tepper, K. (1994). Need for Uniqueness: An individual difference factor affecting
nonconformity in consumer response. In C.W. Park & D.C. Smith (Eds.), Marketing theory
and applications: Proceedings of the 1994 AMA Winter Educators’ Conference. Chicago:
American Marketing Association.
Turner, JC (1982). Towards a cognitive redefinition of the social group. In Tajfel, H. (Ed.),
Social identity and intergroup relations. Cambridge: Cambridge University Press.
Turner, J. C. (1991). Social influence. Pacific Grove, CA: Brooks/Cole.
Divergence 57
Veblen, T. (1899 [1912]). The Theory of the Leisure Class. New York: The Viking Press.
Venkatesan, M. (1966). Experimental Study of consumer behavior conformity and
independence. Journal of Marketing Research, 3, 384-387.
Vincent, L. (2002). Legendary Brands: Unleashing the Power of Storytelling to Create a
Winning Market Strategy. Chicago: Dearborn Trade.
Walker, R. (2004). Yellow fever: What happens when philanthropy and style team up? A charity
craze. New York Times Magazine, August, 9, 23.
Weir, H.B. (1971). Deprivation of the need for uniqueness and some variables moderating its
effects. Unpublished doctoral dissertation, University of Georgia.
Wicklund, R. A., & Gollwitzer, P. M. (1982). Symbolic self-completion. Hillsdale, N.J.:
Lawrence Erlbaum.
Willis, R.H. (1965). Conformity, independence, and anti-conformity. Human Relations, 18, 373-
388.
Wonacott, P. (2004). China's Buick infatuation: The stodgy American auto is a pre-revolutionary
icon for booming middle class. Wall Street Journal, 22 July, B1.
Wong, N.Y., & Ahuvia, A.C. (1998) Personal taste and family face: Luxury consumption in
Confucian and western societies. Psychology & Marketing, 15, 423-441.
Zaltman, G., & Wallendorf, M. (1979) Consumer Behavior: Basic Findings and Management
Implications. New York: John Wiley.
Divergence 58
Figure 1: Divergence in Taste Selection by Taste Domain
65%
25%
10%
Divergence 59
Figure 2: Change in Taste Expression Based on Similarity of the Adopting Group
Divergence 60
Figure 3: Divergence Threshold Based on Similarity of Adopters and Taste Domain
DishSoap
BikeLight
DVDplayer
Color ofPen
CellphoneBrand
FavoriteActor
MusicArtist
MusicGenre
JacketStyle
Hairstyle
Div
erge
nce
Thre
shol
d
Inner City TeensPrinceton Students40 yr-old Biz Execs
Divergence 61
Mathematical Appendix Material There have been a number of important economic theory papers on the subject of fads and fashion. Bernheim (1994) argues that the desire for status is the driver of conformity. Becker and Murphy (2000) argue that leaders abandon old fashions in order to maintain distinction. The contribution of this model is that the driver behind fads and fashion is not status, but identity, where identity is important for allowing effective social interactions. Further, change is not driven by the production technology of the fashion industry (see Pesendorfer 1995), but based on changes in behavior within the population. The richness of the model allows us to identify other properties of fashion, beyond price, that allow us to predict what products and activities are likely to become fashion statements and thus suffer the ups and downs of fad-like behavior. Presented below are the technical details of the models presented in the paper. Let there be a large community of actors of different types--say A and B to start--with population proportion fixed at pA, pB . Presume individuals are myopic to simplify dynamics. Actors have preferences over interactions and over signals. Presume separability, so that in each sub-stage, an agent of type θ who chooses signal i ∈ Z, has history of signal choice h, and interacts with another agent of type θ- with interaction a, receives stage utility of Urecv(a,i|θ,h) = brecvr(a,θ,θ-) - c(i, θ, h) Usend(a,i|θ,h) = bsend(a,θ,θ-) - c(i, θ, h) where b(-) represents the benefits of the interaction and c(-) represents the costs associated with choosing that particular signal. In the interaction stage, one party is randomly selected as sender, and the other as receiver. We take the cost function to be given exogenously. We presume the cost of any given signal is the same for all members of the same type except for an individual specific normally distributed zero-mean error term. Also, each possible signal choice, i, has associated visibility, v(i), the probability it will be observed. Each period of the game proceeds as follows. In the first stage, all actors simultaneously choose a signal. In the second stage they are randomly paired. With equal probability, one is randomly assigned to be the signal sender, the other is the receiver. The receiver observes the sender’s signal chosen in the first stage with probability v(i), and then chooses how to treat the sender. The receiver has preferences over how she treats the sender as a function of their types, and the sender has preferences over how he wants to be treated. Let there be two types of interactions, {α, β}. The receiver always prefers to choose α if the sender is of Type A and always prefers to choose β if the sender is of Type B. That is, all receivers have a preferred mode of interaction with a sender of a given type: brecv(α, A, A) > brecv(β, A, A) and brecv(α, B, A) > brecv(β, B, A) brecv(β, A, B) > brecv(α, A, B ) and brecv(β, B, B) > brecv(α, B, B)
Divergence 62
Similarly, for the preferences of the sender, assume for the moment that individuals of either type are happy with how other types want to treat them. bsend(α, A, A) > bsend(β, A, A) and bsend(α, B, A) > bsend(β, B, A) bsend(β, A, B) > bsend(α, A, B ) and bsend(β, B, B) > bsend(α, B, B) Here, however, for simplicity, assume that all types get utility σ > 0 when they are treated as they wish to be treated, and utility 0 when they are not. That is, let bsend(α, A, A) = bsend(α, B, A) = bsend(β, A, B) = bsend(β, B, B) = σ bsend(β, A, A) = bsend(β, B, B) = bsend(α, A, A ) = bsend(α, B, A) = 0 For the purpose of this analysis, assume costs are history independent. Also, assume for simplicity that if no signal is observed, no interaction occurs and the benefit is zero. We justify this assumption by assuming that the cost of mis-interaction is high in the absence of any information. Solving the Model We focus on pure strategy sub-game perfect Nash equilibria where all individuals of the same type choose the same action. By the myopia assumption, we can focus on just one period. We solve by backward induction; that is, we describe what happens in the second stage of the game where individuals interact, and then use that to solve for what happens in the first stage (the choice of signal). In the second stage, receivers observe the signal and update their beliefs regarding the sender’s type. Based on those beliefs the receiver chooses the appropriate interaction. Beliefs are based on the signals actors choose in equilibrium in the first stage. Specifically, suppose some signal k is observed by the receiver. Let r be the proportion of Type A actors in the population who chose signal k. Let s be the proportion of actors of other types that chose signal k. Then the receiver’s belief that the sender is of Type A is equal to the actual probability the sender is that type and is given by:
srrksignalAtypekm+
==== ]|Pr[)(
We call this expression, m(k), the meaning of signal k. In a world with many people, r and s are ratios set by large populations. The change of signaling behavior by any one person has little impact on the meaning of a signal. Furthermore, we make the assumption that r and s are inferred from observations from past periods, and that off-the-equilibrium-path attempts to signal alone—choosing a signal where m(k)=0—will result in no interaction. The first stage is essentially a coordination game, so nearly any set of signals can be part of a coordination game equilibrium. In this simple example, we have perfect alignment of preferences. If each type chooses a different signal, then there will be perfect identification
Divergence 63
whenever the signal is observed. Thus we would expect Type A to choose a signal i and Type B to choose a different signal j. Utility for each individual is given by the benefits of interaction minus the cost of signaling. Given symmetry, the receiver chooses the action that corresponds to the type that is over 50% likely. Some simple derivations show that the benefits of interaction are as follows: UA = v(i)[ m(i) σ + (1-m(i)) 0] - c(i,A) UB = v(j)[ m(j) σ + (1-m(i)) 0] - c(j,B) However, since in this simple case, each signal identifies the other individual with probability 1, and the receiver will always treat the sender as she wants this simplifies to UA = v(i) σ- c(i,A) and UB = v(j) σ - c(j,B) Thus, the optimization problem is given by maxi ∈ Z UA = v(i) σ - c(i, A) and maxj ∈ Z UB = v(j) σ - c(j, B)
We maximize by differentiating the actor’s utility functions with respect to visibility, vi, and cost, ci.
1−=∂∂
=∂∂
cUv
U
A
A σ
Since the derivative with respect to visibility is positive, actors would choose signals with the highest possible visibility. Since the derivative with respect to costs are negative, actors would choose signals that minimized costs.
Choice of Domains The basic model above can be applied to understand which domains actors choose to signal in. Formally, let us pre-pend another stage before the stages in the game above. Before signals are selected, actors select a domain in which to signal. Define domain to be a set containing each of the possible signaling behaviors actors can choose to engage in. For example, the power saw domain contains the set of possible power saws on the market. The coat domain contains the set of coats one might wear. Each coat has a bundle of attributes each providing a different set of costs and benefits. The baseball cap wearing domain contains the set of possible ways to wear a cap, such as backwards or sideways Assume, for simplicity, that if more than one domain is selected, visibility goes to zero. Actors simultaneously choose a domain in which to signal, effectively determining the choice set of
Divergence 64
signals that are available to them in the signal selection stage. Formally, let D denote the set of possible domains. Then the maximization problem becomes: maxZa ∈ D maxi ∈ Za UA = v(i)[ m(i) σ ] - c(i,A) maxZb ∈ D maxj ∈ Zb UB = v(j)[ m(j) σ ] - c(j,B) In equilibrium, actors will choose the same domain, Zb = Za = Z, since by our assumption miscoordination on domain leads v(-) to go to zero. We see from the utility function that if signaling loses visibility, actors are incurring costs with no benefits. The maximization problem is over the same function, so low cost and high visibility is still preferred. However, here the choice set is over domains. Thus, instead of choosing a signal with one associated cost, the groups are choosing sets of cost. The socially optimal choice maximizes some social welfare function, for instance, the simple sum of utilities Welfare = UA + UB = b(-,A) – c(i,A) + b(-,B) – c(j,B) We saw from the first model that types A and B should optimally choose different signals, i ≠ j, in order to effectively differentiate. Ignoring visibility for the moment, without loss of generality, let i be the signal that minimizes c(i,A) and normalize c(i,A) = c(i,B) = 0. Then, if i also minimizes costs for group B, then signal j can only be second best. i = argmini∈ Z c(i,A) j = argminj∈ Z, j≠ i c(j,B) Since a welfare maximizer cares about the sum of c(i,A) and c(j,B), the second best is crucial. Thus the domains that are best suited for signaling are domains where the difference between the first best and second best are minimized. There are two properties of domains that help minimize this difference. The first is to choose domains where costs are unlikely to be perceived similarly by different groups. In functional domains, the tangible costs and benefits will likely be the same for different groups such that c(s,A) = c(s,B) for all s ∈ Z In these domains, the first best for both groups will be the same, and one will be forced to accept second best. The second is to choose domains where the difference in tangible costs/benefits associated with any particular style are minimal (e.g., those based on color or style). Power saws are mostly differentiated by functional attributes such as power or speed and people are more likely to agree on which saw is optimal; not choosing the optimal saw for a given task is potentially quite costly. However, in the domain of coats, different coats are largely differentiated by color or style and people are less likely to agree on which is optimal: differences in costs and benefits across coats are quite minimal and likely outweighed by differences across people. Thus, the requirement that different groups choose different signals is only minimally costly.
Divergence 65
Though considering the functionality characteristics and cost minimization properties of domains explains a wide variety of domain choice, there are some domains where signaling is quite costly—e.g., body piercings or arcane knowledge—domains whose cost functions defy these predictions. Some other mechanisms must be at work. Our proposed candidate is signal poaching.
Signal Poaching
To consider signal poaching, let us return to considering domains as being exogenously fixed, but now, add a third type, Type BA. Others would prefer to treat BA’s like Bs: brecv(β, A, BA) > brecv(α, A, BA ) and brecv(β, B, BA) > brecv(α, B, BA) However, BA’s would prefer to be treated like A’s. bsend(α, BA, A) > bsend(β, BA, A) and bsend(α, BA, B) > bsend(β, BA, B) Or again, we simplify and say bBA(A, β) = bBA(B, β) = σ and bBA(A, α)= bBA(B, α) = 0 By introducing the BA types, things get more complicated. Type BA would like to choose signal i rather than signal j in the choice of signal stage, but would only do so if the benefits received from the interaction stage outweigh the costs. Let π be the fraction of poseurs that adopt signal i. The fraction π is based on the random noise of the cost function. More precisely, π is the fraction of the poseur population, where the net benefits of posing, v(i) m(i) σ – c(i, BA), are greater than the net benefits of not posing, – c(j, BA).
)],(),()()(Pr[ BAjcBAicimiv −>−= σπ Recall that pA is the fraction of the population of Type A. Note that all actors of Type A will choose signal i (we assume the cost for these types has low enough variance to assure this). In that case, the probability that an actor showing i actually is of Type A would be given as
π+=
A
A
ppim )(
Recall that receivers of Type A will treat the sender as a member of Type A so long as the benefits of choosing α outweigh the benefits of choosing β:
),())(1(),()(),())(1(),()( ββαα BbimAbimBbimAbim AAAA −+>−+ Rearranging we get
Divergence 66
),,(),,(),,(),,(),,(),,()(
BAbBAbAAbAAbBAbBAbim
αβαβαβ
−+−−
>
Rearranging again, and defining the right hand side of the above equation as ρ, we can say that the signal will remain meaningful so long as
⎟⎟⎠
⎞⎜⎜⎝
⎛ −<
ρρπ 1
Ap
Note that ρ gets larger as the penalty from making a mistake increases. Then we can say that a given signal retains meaning so long as π is below some critical threshold that is decreasing in the penalty for a mistake, ρ, and increasing in the proportion of other A types in the population . We can now write sender of Type A’s optimization problem as maxi ∈ Z v(i) [pA/(pA+π)]σ - c(i, A) We can differentiate with respect to π to get
σππ
)()( 2 iv
ppU
A
AA
+−
=∂∂
This term is negative, so Type A wants to keep π as low as possible. One way to do this is by choosing a signal i that makes c(i, BA) as high as possible. However, as before, Type A would still also want to choose a signal i that minimizes costs c(i, A). Combining these two results, we see that in the presence of poaching, the actor is no longer just minimizing costs, but instead, maximizing the difference in costs c(i, BA) – c(i, A), incurred by the poseurs versus the real thing.
When signals are scarce
If we want to consider situations where there are many groups, and only a few signals available, actors have to budget their use of signals. It is reasonable to assume that actors would allocate those signals to applications with the highest payoffs. To extend the model, let us fix one group A, but consider group A’s possible interactions with other groups g, where g could be group B, but could also be group C, group D, etc. each with their own poseurs, of Type BA, CA and DA, etc. respectively. Define fg to be the frequency of interaction with any given group. Now, we can write total utility as
Divergence 67
( )∑ −=g
gAgggA AicgbimivfU ),(),()()( α
Thus, if there were a limited number of signals available, the benefits to signaling would be highest when either frequency of interaction fg was highest, such as the case when dealing with similar others, or when the consequences of failed identification are highest, bA(g,~α) << 0. Alternatively, there could be some exogenous probability of automatic recognition that varies by group dyads. College students from different schools would have a low chance of automatic recognition, whereas teenagers compared to executives would have a high chance of automatic recognition. In cases where automatic recognition is low, the benefits to signaling are high.
Extensions
Clearly this model is far more general than is necessary to explain the phenomena explored in the paper. However, it is intended to provide a systematic general way of thinking about these interactions to allow for future work on questions of cost dynamics, social networks, opportunity costs, interactions with visibility, and a typology of benefit functions. So far we have focused on the interaction and not the identity of counterparty. Imagine, however, the case of a persecuted group (e.g., homosexual men in the early 1960’s). Let this persecuted group be A and the mainstream be B. The persecuted group would like to be treated as mainstream by B receivers, but be treated as themselves by A members: bsend(α, A, A) > bsend(β, A, A) and bsend(β, A, B) > bsend(α, A, B) Thus they would choose signals visible to their group, but hidden to the mainstream.
